An engine for ML/data tracking, visualization, explainability, drift detection, and dashboards, integrated with Polyaxon.
TraceML is an observability engine for machine learning and data science that provides experiment tracking, visualization, explainability, and drift detection. It integrates with Polyaxon to help teams log, monitor, and analyze ML experiments and data workflows. The tool supports offline usage and offers extended data profiling capabilities for pandas DataFrames.
Data scientists and ML engineers using Polyaxon for MLOps who need robust experiment tracking, visualization, and data profiling within their workflows.
Developers choose TraceML for its deep integration with Polyaxon and major ML frameworks, its rich visualization support across multiple libraries, and its comprehensive data profiling features that extend standard pandas summaries.
Engine for ML/Data tracking, visualization, explainability, drift detection, and dashboards for Polyaxon.
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Offers native callbacks and loggers for Keras, PyTorch, TensorFlow, Fastai, PyTorch Lightning, and HuggingFace, simplifying integration across popular ML libraries as shown in the README examples.
Supports multiple plotting libraries including Matplotlib, Plotly, Altair, and Bokeh, allowing flexible and interactive visual insights directly from tracking code.
Enables local tracking without a Polyaxon API server via POLYAXON_OFFLINE environment variable or is_offline flag, useful for development and testing.
Provides DataFrameSummary with detailed statistics beyond pandas describe(), such as column types, missing percentages, and correlation insights for better data understanding.
Core tracking features require polyaxon installation, making it less suitable for teams not committed to the Polyaxon ecosystem and adding vendor lock-in.
Key components like local sandbox, summary visualizations, and catalog integration are marked as WIP in the README, limiting current functionality and reliability.
Initial configuration involves specifying projects, artifacts paths, and tags, which can be cumbersome for quick experiments compared to simpler logging tools.